The interaction of the trp repressor from Escherichia coli with the trp operator

Biochimica et Biophysica A cta 909 (1987) 58-70 Elsevier

58

BBA 91719

T h e i n t e r a c t i o n o f t h e trp r e p r e s s o r f r o m Escherichia coli with t h e trp o p e r a t o r

A n d r e w N. L a n e , J e a n - F r a n c o i s L e f ~ v r e * a n d Oleg J a r d e t z k y Stanford Magnetic Resonance Laboratory, Stanford University, Stanford, CA (U.S.A.) (Received 4 December 1986)

Key words: Repressor-operator interaction; Imino proton exchange; NMR; Protein-DNA interaction; Tryptophan operator

We have examined the interaction of the trp repressor from Escherichia coli with a 20 base-pair synthetic operator. Nonspecific binding was relatively strong (Ko-----2 pM), but only weakly sensitive to the concentration of added salt ((d log K d ) / ( d log [Na]) = - 1). 1H-NMR studies indicate that the structure of the repressor is not greatly altered on forming the complex, and that few if any of the lysine and arginine residues make direct contact with the DNA. However, the mobility of one of the two tyrosine residues is significantly decreased in the complex. The repressor makes close contact with the major grooves of the operator such that the base protons are broadened much more than expected on the basis of increased correlation time. There are large, differential changes in chemical shifts of the imino protons on forming the complex, as well as changes in the rate constants for exchange. The fraying of the ends is greatly diminished, consistent with a target size of about 20 base-pairs. The effects of the repressor on the NMR spectra and relaxation rate constants can be interpreted as a change in the conformation of the operator, possibly a kinking in the centre of the molecule.

Introduction

An important aspect of the control of transcription in bacteria is the recognition and binding of regulatory proteins to special sequences of DNA, such as operators and promoters. It has been argued on the basis of conservation of the sequences of target sites and model building that the specificity of the interaction arises largely from specific pairing of hydrogen bond donors and acceptors in the protein and the DNA [1]. While hydrogen bonding patterns may account for some of the specificity (Lef6vre, J.-F., unpublished data), * Present address: Institut de Biologie Moleculaire et Cellulaire, Strasbourg, France. Abbreviation: DSS, 2,2'-dimethylsilapentane 5-sulphonate. Correspondence (present address): A.N. Lane, National Institute for Medical Research, The Ridgeway, Mill Hill, London, NW7 1AA, U.K.

other factors are also important [2]. Using the X-ray structure of the EcoRI restriction site, Calladine [3] and Dickerson [4] have proposed a set of rules for estimating the helical twist, roll and puckers of any sequence of DNA. There are indications that these rules also apply to DNA in solution [5,6]. Further, application of these rules to long stretches of DNA containing operators and promoters indicates that the control sites have structures significantly different from the bulk DNA [7]. It is likely, therefore, that some of the sequence specificity of binding arises from recognition of geometric features. On the other hand, simple application of these rules does not give any indication of regions of a sequence that are more flexible than the average. We have observed a flexible region in the trp operator [8]. However, most model building of repressoroperator interactions has assumed standard B-

0167-4781/87/$03.50 © 1987 Elsevier Science Publishers B.V. (Biomedical Division)

59 D N A geometry, and rigid bodies [1,9]. The results of Dickerson's and our own studies [4,6,8] show that the former assumption is unjustified. Also, calculations on the c r o - O R 1 interaction, and the cAMP-binding protein and its target site are consistent with significant bending of the D N A [10,11]. More conclusive is the X-ray diffraction of the complex of the E c o R I restriction endonuclease with its target site, in which the backbone of the D N A clearly kinks [12,13]. Hence, to understand the details of specific p r o t e i n - D N A interactions, one must take into account not only the actual structures of the components, rather than assumed ones, but also their intrinsic conformational flexibilities. To this end we have been studying the structures and properties of the trp repressor from E. coli [14-17] and the trp operator [6,8,18]. The X-ray structure of the trp holorepressor has been determined [9]. We have also investigated the effect of the trp repressor on the structure and dynamics of dA20dT20, which represents a wholly nonspecific interaction [19]. In this article, we report the thermodynamics of the trp repressor-operator interaction and the influence of the repressor on the structure and properties of the operator, comparing the results with the nonspecific interaction. Materials and Methods Materials

The trp o p e r a t o r / p r o m o t e r (sequence: CGTAC T A G T T . A A C T A G T A C G ) was prepared as previously described [18]. The trp repressor was prepared from an overproducing mutant of E. coli, and assayed as described [14]. All other reagents were of the highest purity commercially available. Methods

Circular dichroism spectra were recorded on a JASCO spectropolarimeter using a 1 cm pathlength. Titrations were performed by adding small aliquots of a concentrated repressor solution to a solution of the operator, and subtracting the intrinsic ellipticity of the repressor. The sensitivity of binding to the concentration of sodium chloride was established by adding small aliquots of concentrated, buffered sodium chloride solution to a solution of preformed complex, and monitoring

the decrease in elhpticity at 276 nm [20]. 1H-NMR spectra were recorded at 500 MHz on a JEOL GX-500 spectrometer, using DSS as an internal reference. Solutions of repressor and operator were separately lyophilised and redissolved in 100% 2H20. Their concentrations were determined from absorbance at 260 and 280 nm, respectively. The solutions were diluted with 2H 20 to give identical concentrations. The complex was prepared by mixing equal volumes of the two solutions (final concentration = 0.9 mM in duplex and 0.9 mM in repressor dimer). The reference solutions of repressor and operator were therefore at twice their concentrations in the complex. The pEH* was 8.5. J-resolved spectra were obtained with the 90-~'180-~--Acq pulse sequence [21]. Selective spinlattice relaxation times and nuclear Overhauser effects were measured as previously described [18]. To form the specific complex, tryptophan in 2H20 was added in small increments until saturation (i.e., 1.8 raM). Difference spectra were calculated from transformed, phased spectra to which 5 Hz line-broadening was added. After all the experiments on the non-exchangeable protons were done, the solution of the complex was lyophilised, and redissolved in 85% XH20/15% 2H20. The pH was adjusted to 6.8 with 2HC1, and to 9.6 with NaO2H. Controls showed that lyophilisation had no deleterious effects on the properties of the macromolecules (for example, the repressor was active in the assay, and the N M R spectrum was not significantly altered). The assignments of the imino protons were checked by comparing areas of the peaks, chemical shifts with those in the free operator, and interimino proton nuclear Overhauser effects. Nuclear Overhauser effects were measured according to Wagner and Wiithrich [22]. The rate constants for exchange of the imino protons with solvent were measured as a function of temperature and p H by saturation recovery as previously described [18]. The magnetisation recovery curves were fitted by non-linear regression to the equation: M ( t ) = M °¢ - ( M o + M °°) exp[- Rlt ]

(1)

where M ( t ) is the magnetisation at time t after

60 the saturation pulse, M o and M °o are the initial and equilibrium magnetisations, respectively, and R 1 is the observed spin-lattice relaxation rate constant. N M R spectra were calculated as the sums of Lorentzians using the data previously obtained [6,14] and an interactive program written for an Apple Macintosh (Lane, A.N. and Higashi, R.M., unpublished data). Results and Discussion

Thermodynamics of the interaction of the trp repressor with the trp operator To determine whether the trp repressor is a double-stranded D N A binding protein, we estimated its effect on the melting temperature of the operator. In 0.1 M phosphate buffer (pH 7.5) the melting temperature of the operator fragment determined optically at 260 nm was 56°C. In the presence of a slight excess of repressor and tryptophan ( K d ---1-5 nM), the melting temperature increased by 5 Cdeg, indicating that the holorepressor stabilises the double-stranded state. We have shown that the trp repressor maintains its tertiary structure under these conditions up to at least 70 ° C [14]. The apparent van't Hoff enthalpy for melting in the absence of the repressor was 90 kcal/mol, and 80 k c a l / m o l in its presence. To a first approximation, the difference in enthalpy is equal to the enthalpy of binding of the holorepressor to the operator. This is because the concentration of free repressor at a total concentration of 10 /~M is much larger than the dissociation constant (about 5 nM [23]). Hence, the binding enthalpy is approx. - 1 0 kcal/mol. This should be compared with the binding enthalpy of the lac repressor with the lac operator ( - 6 kcal/mol), and the nonspecific binding of the trp repressor to dA 20dT20 (about - 1 0 k c a l / m o l [19]). The dissociation constant for the repressor-operator complex has been estimated at about 5 nM in the presence of tryptophan [23,24], and shown to be much less in the absence of tryptophan, though no estimate of the dissociation constant was given. We therefore sought to determine the affinity in the absence of tryptophan. Fig. 1A shows CD spectra of the operator in the presence and absence of the repressor. The CD of the

repressor (curve 1) is weak compared with that of the free operator (curve 2), whose spectrum is typical of D N A in the B family of conformations. Addition of the aporepressor results in a significant increase in ellipticity (about 50% at 100% saturation), with no major shift in the wavelength (curves 3 and 4). Titration of the CD signal with increasing concentration of the aporepressor leads to curve, that displays a break when saturation is reached (Fig. 1B). The break occurs at a concentration of the aporepressor dimer equal to the concentration of the duplex. This indicates that one repressor dimer binds per 20-mer, and that the target size is of the order 20 base-pairs or more. We note that a 20-mer is equivalent to two turns of the double helix, just sufficient to interact with a protein dimer whose length is also approx. 68 .~ [9]. The titration curve in Fig. 1B can be analysed to yield a dissociation constant of 2 /~M at 25 o C. This affinity is substantially higher than for binding to dA20dT20, where K d is approx. 70 /~M in the presence of only 10 mM sodium chloride [19]. In the presence of excess tryptophan, the binding curve becomes stoichiometric at the lowest concentrations of D N A where a signal is still observable, implying a dissociation constant for the holorepressor of much less than 0.2 ttM. Nevertheless, this result shows that the mutant operator we are studying responds normally to the repressor. The thermodynamic constants for the interaction of the trp repressor with DNA are listed in Table I. We have shown also that the binding of the trp repressor to dA20dT20 is only weakly dependent on the concentration of sodium chloride [19]. If aliquots of concentrated solutions of sodium chloride are added to the aporepressor-operator complex, the CD signal decreases. Hence, the apparent dissociation constant as a function of the concentration of added salt can be estimated. The results are plotted in Fig. 1C. The slope of the line is about - 1 . The affinity of the trp repressor for D N A is much less sensitive to ionic strength than is the affinity of the lac or cro repressors for their operators [20,25]. A comparison of some of the thermodynamic parameters for repressor-operator interactions is given in Table I. The most striking differences between the interaction of the trp repressor with DNA and other repressor-DNA in-

61 A

B 4

5 ~

/ /

-

-

i

0

4~

I

-4

12

8

Repressor c o n c e n t r a t i o n

(~jM)

(2

1.2

0.4

I 240

i

I 280 ). (rim)

i

l 320

i -Log

1'.0 ~oCI]

i

21.0

Fig. 1. Interaction of the trp repressor with the trp operator fragment monitored by circular dichroism. The concentration of DNA was 1.6/tM in 20-mer, in 0.1 M phosphate buffer (pH 7.5), 25 o C. (A) near-ultraviolet CD spectra of the operator, the repressor and the operator-repressor complex. The concentration of repressor was 3.9/tM in dimer. 1, repressor+ 30 /~M L-tryptophan; 2, free operator; 3, operator + aporepressor; 4, operator+ aporepressor+ 30 /~M E-tryptophan. (B) CD titration monitored at 276 nm. The dotted line shows the increase in ellipticity due to addition of aporepressor. The continuous curve is a fit to the data for a dissociation constant of 2 /~M. (C) Dependence of the apparent dissociation constant on the concentration of added NaCI. The dissociation constant was estimated as described in the Materials and Methods section. teractions is the low sensitivity of the binding to the concentration of sodium chloride. However, the interpretation of the effects of ionic strength on affinity in terms of ionic contacts is not necessarily valid, as what counts is the distribution of charges on the surface of the protein and the global electrostatic potential. In fact, the repressor bears a nett negative charge at physiological pH [24]; this would lead to the expectation of increasing affinity with increasing ionic strength in a point charge model. Nevertheless, the weakness of the dependence of the affinity on the concentration of salt suggests that interactions other than purely electrostatic are important for the trp repressor.

Interaction of the trp repressor with the trp operator observed by IH.NMR N M R spectra. In the absence of conformational

changes, the interaction of the repressor and the operator fragment can be expected to increase the linewidths in proportion to the fractional increase in rotational correlation time (see below). The correlation time for the operator (Mr = 12000) should increase nearly 3-fold, and the repressor ( M r = 25 000) about 50% on forming the complex ( M r = 37000). Changes other than in the linewidths or chemical shifts have then to be attributed to three possible causes: (1) change in conformation; (2) effects of new nearest neighbours at the interface of the complex; and (3) changes in internal motions. Fig. 2 compares simulated spectra with the observed spectra of the components and the complex, assuming only effects on linewidths proportional to the increased correlation times. The simulated spectra adequately represent the observed spectra of the individual components, but

62

B

B'

~A c

,

85

i

.

i

• 5

,

,

65

.

i

55

6(ppm)

6(ppm) Fig. 2. Interaction of the trp repressor with the operator monitored by 1H-NMR. trp operator was 0.9 mM in 20-mer and the repressor 0.9 mM in dimer in phosphate buffer (p2H * 7.5), 25 ° C. Spectra were simulated as described in the Materials and Methods. (A) Aromatic spectrum of the trp repressor. The asterisks denote incompletely exchanged amide protons. (A') Simulated spectrum of the trp repressor. (B) Low field spectrum of the trp operator. (B') Simulated spectrum of the trp operator. (C) Spectrum of the trp operator-repressor complex. ( C ' ) Simulated spectrum of the complex assuming line-broadening proportional to the increase in rotational correlation times of the components (3-fold for the D N A and 1.5-fold for the protein).

TABLE I COMPARISON OF T H E R M O D Y N A M I C PARAMETERS FOR T H E I N T E R A C T I O N OF D I F F E R E N T REPRESSORS WITH DNA OR3(9) and OR3(17) are 9- and 17-mers, respectively. K d is the dissociation constant, A H is the van't Hoff enthalpy change, 1 is the ionic strength, and (d log K d ) / ( d log [Na]) gives the dependence of the dissociation constant to added sodium chloride. Repressor

lac lac lac cro ~, cro trp trp + W trp

DNA

lac O CT a dAdT OR3(9) OR3(17) trp O trp O dA 2odT2o

a Calf thymus DNA. b n.d., not determined.

Kd

AH

Condition

(#M)

(kcal/mol)

t ( ° C)

I (M)

d log K a

10- 8 1 2 19 0.03 2 < 0.1 75

8 - 6 n.d. b n.d. n.d. - 10 n.d. < - 10

25 25 25 8 8 25 25 20

0.05 0.1 0.03 0.1 0.1 0.1 0.1 0.1

Ref.

d log [Na] - 6 - 9 n.d. - 7.6 - 12.6 - 1 n.d. - 2

25 25 25 20 20 this work this work 19

63 not at all well that of the complex. In particular, the H8 protons of the bases are broadened almost beyond detection, and the H I ' region is also broadened more than expected. We have previously shown that the bases in the free operator do not undergo significant internal motions [26], so that quenching of internal motions of the bases in the complex is unlikely to account for the excess line-broadening. However, the H8's are in the major groove, so that interaction with groups on the surface of the repressor could provide additional relaxation pathways, and therefore increase the linewidths more than the expected 3fold. An alternative mechanism is that scalar relaxation from the two laN nuclei nearby in the ring is increased in the complex. This mechanism is not very efficient for the H8's, as the two-bond coupling constant is small. Further, the relaxation rate is inversely proportional to the correlation time, so this mechanism will contribute less to relaxation in the complex than in the free DNA, unless the quadrupolar coupling constant increases proportionally more than the increase in the correlation time. Although the enhanced scalar relaxation cannot be definitely ruled out, it is less likely than enhanced dipolar relaxation due to additional nearest neighbours from the protein. The H l ' s , on the other hand, are relatively sharp in the free D N A for two reasons, namely there is a relatively small number of close protons, and the correlation time is short owing to pseudorotation. It is possible that the interaction affects the rate or the amplitude of the pseudorotation. Because the lines are so broad at 25 o C, it is difficult to measure their properties. We therefore raised the temperature to 40°C, which is well below the melting temperature of either component (see above). This is expected to decrease the correlation time, and therefore the linewidths, by about 30% [26]. Interactions between side-chains of the protein with the D N A are expected to have a variety of possible effects, including changes in chemical shifts, (especially where hydrogen bonding or salt bridges are concerned), differential changes in relaxation rate constants, and changes in coupling constants. Spectra recorded at 40 ° C (not shown) have significantly better resolution than those recorded at 25°C. With the aid of resolution enhancement, it is possible to pair the

resonances with those in the isolated components (also at 40°C), suggesting that, in the majority of cases, changes in chemical shifts are small. A convenient method of observing changes in chemical shifts and coupling constants is the twodimensional J-resolved experiment [21]. In this experiment, the chemical shifts are separated from the splitting due to scalar coupling, and is useful for examining the coupling patterns of resolved residues. In the region 2.5 to 3.2 ppm, only the protons of lysine, the 8 protons of arginine and fl protons of aromatic residues resonate. The lysine c protons, which are sensitive to changes in pH [14,16], are found at 2.6 to 3.15 ppm. Addition of the trp repressor to the trp operator has no significant effects either on the chemical shifts (AS < 0.03) ppm) or on the coupling patterns (cf. Table III). Relaxation rates. Another N M R parameter that is sensitive to interactions is the intrinsic spinlattice relaxation rate constant, which is sensitive to internal motions. The intrinsic spin-lattice relaxation rate constant, Pl, is related to the correlation time, ~', by [26,27]: pl=(vah2/40~r2)[J(tol-~2)+3J(~l)+6J(~l+to2)]

(2)

where ~, is the gyromagnetic ratio, h is Planck's constant, and J(to) are the spectral density functions. For macromolecules, the spectral density function J(to 1 - o:2) dominates the relaxation rate, and is given approximately by: J ( 6 0 1 - 602) = "r

(3)

Apart from small changes in the shape of the molecules on forming the complex, the correlation time i s proportional to the molecular weight. Hence, the formation of a complex between two rigid macromolecules would be expected to increase the correlation times of the components in proportion to the increase in the molecular weight. This assumes that there are no drastic changes in shape. Measurements of correlation time for internuclear vectors in the D N A moiety are consistent with the expected 3-fold increase in the molecular weight [19]. Measurements of the intrinsic spinlattice relaxation rate constant for the lysyl c-CH 2 resonances are consistent with only an overall increase in correlation time due to binding the

64

DNA. In the presence of significant interaction of the ammonium group, one would expect a much larger increase in the relaxation rate, as these side-chains have significant freedom of motion with respect to the protein [16]. Similar remarks apply to the arginine resonances. Together with the absence of an effect of the repressor on the chemical shifts and coupling patterns (see above), these data support the finding that the affinity of the repressor for D N A is only weakly dependent on the concentration of added salt, implying that few salt bridges are formed. We have measured the intrinsic spin-lattice relaxation rate constants of other resolved resonances at 40 o C. The values are given in Tables II and III. The values of the relaxation rate constants for protein resonances are up to 50% higher in the complex than in the the free repressor, consistent with an increase in the overall tumbling time (see above), and increases in line widths of 50% or less. The exception is that of a tyrosine resonance (~ = 6.85" ppm), whose relaxation rate constant increases 2.3-fold in the complex. This tyrosine is known to undergo significant internal motion in the free repressor [16]; its motion may be somewhat quenched in the complex. However, the tyrosine residue does not intercalate into the DNA, as the change of the chemical shift is only 0.02 ppm. Indeed, there is no evidence for intercalation of any aromatic residues into the DNA. This is consistent with other repressors binding to double-stranded D N A [27].

S P I N - L A T T I C E R E L A X A T I O N R A T E C O N S T A N T S IN T H E trp O P E R A T O R A N D T H E R E P R E S S O R - O P E R A T O R COMPLEX Intrinsic spin-lattice relaxation rate constants were measured as described in the Materials and Methods. Pu is purine, Py is pyrimidine; n.d., not determined. Assignment

(ppm) 6.2 6.1 5.83 5.67 5.44 5.25

Pu H I ' Pu H I ' HI' Py H I ' Py H I ' C5H5

A

Compel×72,5,6 --815 go J5 Jo 6'5 do

ppm Repersor ~ ~Compelx

~'~ go 4~ 7'0 d5 ~o

~

B

86 84 d2 dO 82 I 0'6 84 82 C~0-C~2 Chemicalshift(Dprn)

T A B L E II

8

The relaxation rate constants for the H l ' s in the complex are surprisingly large. In the free operator, the relaxation rate constants at 4 0 ° C

R 1 (s - l ) free

complex

2.1 1.2 1.6 2.0 n.d. 1.6

n.d. 3.2 9.6 9.6 11 5.9

R 1 (complex) R l (free) 2.7 6.0 4.8 3.7

Fig. 3. Interaction of the complex with tryptophan. Conditions were as in Fig. 3. 2000 F I D s consisting of 8192 points over a spectral width of 5000 Hz were acquired for each spectrum. Buffered tryptophan solution was added in small aliquots until saturation was reached. For the difference spectra, individual spectra were apodised with a 4 Hz line-broadening exponential before subtraction. The spectra in the high field region were mildly resolution enhanced with a Lorentz to Gaussian transformation. (A) Difference spectra in the aromatic region. Upper spectrum, complex + tryptophan-complex; lower spectrum, aporepressor+tryptophan-aporepressor. (B) High-field region as a function of the fractional saturation with tryptophan. Left spectra, repressor + tryptophan at fractional saturations of 0, 0.35 and 0.82. Right spectra, c o m p l e x + t r y p t o p h a n at fractional saturations of 0, 0.4 and 1.0. The arrows point to a methyl resonance that responds to the ring current of the b o u n d tryptophan [17].

65 TABLE III E F F E C T O F T H E F O R M I N G T H E COMPLEX ON T H E PROPERTIES OF T H E trp REPRESSOR AT 40 ° C Coupling constants were estimated from J-resolved experiments, and chemical shifts from resolution enhanced spectra at 40 o C. Pi is the intrinsic spin-lattice relaxation rate constant determined as described in the Materials and Methods. CPX is the operator-repressor complex. D, doublet; T, triplet; D / D , double doublet. Proton

8(free) (ppm)

8(CPX) (ppm)

J(free) (Hz)

J(CPX) (Hz)

pl(free) ( s - 1 )

pl(CPX ) ( s - l )

Y5,6 F22 F22 Arfl K¢ Kc Kc Kc n.a. b V Me

6.85 6.4 6.3 3.26 3.13 3.02 2.98 2.64 2.56 2.40 0.32

6.87 6.4 6.3 3.23 3.15 3.01 3.01 2.67 2.58 2.40 0.30

7.5(D) 8(T) ll(D) 8(T) 10(T) 7,13(D/D) 9(T) 10(T) -

7.5(D) 9(T) 12(D) 9(T) 10(T) ll(T) ll(T) 10(T) -

1.1 10 11 3.3 n.d. " 2.0 2.0 n.d. n.d. n.d. 44 c

2.5 14 13.5 2.4 n.d. 2.3 2.3 n.d. n.d. n.d. 63 c

L Me

0.04

0.024

63 c

119 c

RS/Arfl

-

-

a n.d., not determined. b n.a., not assigned. c value of R 2 determined from linewidths.

are about 2 s-1 (cf, Table II). For a 3-fold in the correlation time of the DNA moiety in the complex, one would expect a 3-fold increase of the relaxation rate constant. The observed increases are substantially greater (cf. Table II). This agrees with the additional line broadening observed for these protons (see above). Either groups on the protein approach the H l ' s close enough to provide additional relaxation, or there is a conformational effect such that the amplitude or frequency of the pseudorotation is decreased (leading to a greater than 3-fold increase in the effective correlation time). The chemical shift of the H5 of C5 is not significantly perturbed in the complex (AS < 0.02 ppm), its linewidth is greatly increased, and the spin-lattice relaxation rate constant increases from 1.6 to 5.9 s -1, a factor of 3.5. This increase reflects predominantly the increased rotational correlation time of the DNA in the complex.

Interaction of the aporepressor-operator complex with L-tryptophan The binding of the corepressor, L-tryptophan, and an inducer, indolepropanoic acid, to the aporepressor has been studied by a variety of spectroscopic techniques which show that there

are no major changes in conformation of the protein. The binding site is of low polarity, and the amino group in the ammonium form makes an important interaction necessary for function [17]. The X-ray structure [9] shows that an arginine residue probably makes a salt contact with the carboxylate, and the amino group seems to interact with the C-terminus of one of the helices. Fig. 3A shows NMR difference spectra in the aromatic region for tryptophan binding in the absence and presence of the operator. The difference spectra are very similar, being dominated by the peaks of the bound ligand, and perturbed protein residues (at 7 ppm). In the 5 to 6.5 ppm region, where only Phe-22 in the protein resonates, there are difference peaks corresponding to changes in chemical shifts and linewidths of H l ' s of the DNA. This may indicate some further rearrangement in the DNA moiety in forming the specific complex from the nonspecific complex. However, the similarity of the ligand-bound spectrum in the absence and presence of the operator is consistent with a similar mode of binding, and similar properties of the binding site. The aliphatic region is too crowded to interpret at present, being filled with many small difference peaks. However, in the high-field region of the

66

spectrum (Fig. 3B), a new peak is observed at 0.44 ppm, which is identical to that observed in the absence of the operator [17]. The arrow points to the ring-current shifted methyl resonance induced by the ligand [17]. The similarity of the shift in the absence and presence of the DNA strongly argues that the relative orientation of the indole moiety of the ligand to the methyl group is similar in both cases. Evidently, the D N A does not significantly affect the relative orientation of the methyl group of Val-58 with respect to the plane of the ring [9,17]. As far as tryptophan binding is concerned, the changes in structure of the repressor in specific versus nonspecific complexes are subtle. A

i

Operotor~j~~

~2

1~o ld8 ~6

Chem,col shift (ppm)

.~

i

|

i

i

(

>, ~ , O - O --, ~d

o

6o

4.

2&o 3&o %me (ms)

Fig. Effect of the trp repressor on the imino protons. The complex was formed from 0.9 m M each of operator and r e p r e s s o r + l . 8 m M tryptophan in 85% 1 H 2 0 : 15% 2 H 2 0 (pH 6.8). (A) Upper spectrum, imino region of the operator-repressor-tryptophan complex; lower spectrum, imino region of the free operator. 2000 acquisitions for a sweep width of 12000 Hz using 16384 points. The spectra were apodised with a 4 Hz line-broadening exponential. (B) Saturation recovery curve of the GC5, 8 and AT7 imino protons at 2 5 ° C (pH 6.8). The lines are the non-linear least-squares fit to Eqn. 1 as described in Materials and Methods (O) G C 5 / 8 (12.6 ppm); ( O ) AT7 + TA6 (13.6 ppm).

The effect of the trp holorepressor on the stability of the imino protons We have previously assigned the imino protons in the free operator, and have measured the exchange rate constants as a function of pH and temperature [17]. We demonstrated that the rate constants depend on the sequence through differences in the apparent entropy of activation rather than the enthalpy of activation, which was suggested by several authors [28,29]. Because the rate constant for exchange measures a structural fluctuation in the backbone leading to rupture of the Watson-Crick hydrogen bonds [30], we are interested in the effect the trp repressor has on these fluctuations. We have previously shown that in the nonspecific complex with dA20dT20, the exchange constant decreases, with an increase in both the activation enthalpy and the activation entropy [19]. Fig. 4 shows the imino proton region of the operator in the absence and presence of holorepressor. As expected, the resonance linewidths in the complex are substantially larger than in the free DNA. Taking the resonance at 12.6 ppm to represent two protons (GC5/8), then the spectral region consists of ten protons (at 20°C, pH 6.8, where exchange is negligible). Further, control experiments with the free repressor or complexed with tryptophan showed no resonances downfield of 10 ppm [17]. Hence, all of the imino protons are present, with no contamination from protons of the protein. It is clear that the intense peak at 12.6 ppm is from the G C base-pairs 5 and 8 by comparison with the free operator, and the low-field peak at 14 ppm arises from TA9. However, given these assignments, the G C 5 / 8 protons do not shift significantly on forming the complex, whereas that of TA9 moves 0.26 ppm downfield. Therefore, there may be differential effects on the structure of the DNA. Further, in the complex, there are two protons in the region 13.0 to 13.2 ppm, where in the complex there was only 1, from GC1. Hence, to establish the assignments, we performed the same experiments as before [18]. Because of the lower concentration, and broader lines, the signalto-noise ratios in nuclear Overhauser difference spectra were somewhat poorer, but nuclear Overhauser effects from G C 5 / 8 clearly established the

67 TABLE IV A S S I G N M E N T S OF T H E I M I N O PROTONS IN T H E trp HOLOREPRESSOR-trp OPERATOR COMPLEX The resonances in the complex at 20 o C are lettered from A to H from low to high field (cf. Fig. 3A). The chemical shifts were determined from resolution enhanced spectra, n . is the number of protons estimated from areas of each resonance, assumhag that resonance H corresponds to two protons. The area of resonances B + C and C + D each correspond to two protons. Nuclear Overhauser effects (NOEs) were measured as previously described [18] using an irradiation time of 100 ms. The assignment of resonances B and C is ambiguous owing to the overlap. Resonance

8(ppm)

nr~

NOEs to:

Assignment

A B C D E F G H

13.98 13.62 13.52 13.44 13.20 13.08 12.72 12.58

1 1 1 2 1 1 1 2

E,H H a H a n.i. b A,B+C c A,B+C ¢ n.i. A,B + C,E

TA9 AT7/TA6 TA6/AT7 A T 4 + TA3 TA10 CG1 CG2 CG5 + GC8

a These two resonances were simultaneously irradiated. b n.J., not irradiated. These two resonances were simultaneously irradiated.

assignments of AT4, TA6, and AT7. The peak at 12.7 p p m we assign to a G C proton on the basis of its chemical shift, and by comparison with the spectrum of the free operator, it is CG2. Similarly, CG1 is found at 13 ppm. This leaves TA3 and TA10 to be assigned. Irradiation of the resonance at 14 p p m (TA9) gave nuclear Overhauser effects to the peaks at 12.6 p p m (GC 5, 8) and to the peak at 13.2 ppm. This establishes the peak at 13.2 p p m as TA10, and therefore TA3 is in the broad peak at 13.4 ppm. The imino proton of TA10 therefore shifts 0.24 p p m upfield. This is equal and opposite in sign to the shift of the neighbouring imino proton of TA9. Table IV summarises the assignments of the imino protons in the holorepressor-operator complex. As we had previously demonstrated that, in order to be sure that exchange rate constants are true measures of the structural fluctuation in the DNA, it is necessary to make the measurements at high p H [18]. We have therefore measured selective spin-lattice relaxation rate constants at p H 6.8 and 9.6 as a function of temperature. The rate

constants for exchange were then calculated by subtracting out the magnetic contribution as previously described [18]. Fig. 6B shows typical saturation recovery curves, and the best fit lines through the points. The results are shown in Table V. Because of limited signal-to-noise ratios, relaxation rates at relatively few temperatures were measured. Indeed, only the resonance of G C 5 / 8 retained sufficient intensity at 40 ° C to make the measurements. The data at p H 6.8 show that there is insignificant exchange at 2 0 ° C under these conditions, therefore providing a baseline for the data obtained at p H 9.6. To facilitate comparison with the rate constants in the free operator, the values at 35 o C are compared in Table VI. The rate constant for exchange of the imino proton of GC1 is decreased by at least a factor of 60 at 35°C, and that of GC2 and TA3 about 25-fold. These three protons clearly demonstrate fraying of the ends in the free operator. This shows that the repressor greatly diminishes fraying of the ends, consistent with the protein covering most if not all of the 20 base-pair fragment (and see above). The rate constants for AT4, TA6, and AT7 on the other hand are barely affected at 35°C, and are also similar at 25°C, implying similar activation parameters in the free and bound states. The rate constant for TA9 is about 3-fold higher in the complex at 35 ° C, and the apparent activation energy is about 15 kcal/mol. This value m a y not be significantly different from that in the free operator, given the experimental precision, implying that the activation entropy increases in the complex. The rate constant for G C 5 / 8 increases about 3-fold in the complex, the activation enthalpy increases from 18 to 25 k c a l / m o l , and the activation entropy from 3.5 to 29 c a l / m o l per K. Hence, the increase in the activation entropy more than offsets the increase in the activation enthalpy. One can define a critical temperature at which the rate constant in the complex is equal to the rate constant in the free operator, Tc = ( A H* (complex) -- A H* (free)) / ( zaS* (complex) - A S* (free))

For G C 5 / 8 , Tc is about 286 K. Above this temperature, the rate constant in the complex is greater

68 TABLE V R E L A X A T I O N A N D E X C H A N G E C O N S T A N T S F O R T H E trp O P E R A T O R - R E P R E S S O R - T R Y P T O P H A N C O M P L E X A T p H 6.8 A N D p H 9.6 Relaxation rate constants were estimated from saturation recovery curves at different temperatures. R 2° is the relaxation rate constant at 20 o C. Quoted values are best-fit estimates by non-linear regression to the equation M ( t ) = M ~ + ( M o - M ~)exp( - R 1t). Values in parentheses are estimates of the exchange rate constant. Chemical shifts are given at 21 o C. Assignments are from Table IV. kx is the exchange rate constant calculated from R 1 as described [18]. n.d., not determined. 1. p H 6.8 8(ppm) assign

12.58 GC5/8

13.08 CG1

13.2 TA10

13.6 TA6/7

13.98 TA9

R~ ° (s - 1 ) R ~ (s - 1 )

12.6 9.7

14.6 14

14.6 14

n.d. 13.6

12.6 n.d.

2. p H 9.6 8

Assign

(ppm) 12.58 12.72 13.08 13.2 13.44 13.52 13.65 13.98

Rl(kx)(s -1) t ( o C): 10

GC5/8 GC2 GC1 TA10 AT4/TA3 TA6 AT7 TA9

13.6(0) n.d. 16(0) 16(0) 25(8) 18(0) 14(0) 8(0)

20

25

30

35

40

15.8(3.2) n.d. 14(2.4) 15(3) n.d. 14(1.4) n.d. 18(6)

15.8(4.8) 20(9) 25(14) 22(11) 55(43) 35(23) 25(16) n.d.

16.9(7.2) 30(20) n.d. n.d. 60(49) n.d. 60(52) 24(14)

33.0(24.4) n.d. 25(17) 30(22) 150(140) 45(36) 55(48) 50(43)

55(47) n.d. n.d. n.d. n.d. n.d. n.d. n.d.

than in the free operator, and the reaction becomes entropy-controlled. In the complex with dA 20dT20, the critical temperature is above 400 K, so that at all accessible temperatures, the rate constant is smaller in the complex than in the free DNA [19]. Interestingly, the chemical shifts of the GC

protons are not greatly affected by the repressor, though the temperature-dependence of GC2 is. There is a gradation of changes in chemical shifts of the AT protons, with the largest being for TA9 and TA10, in opposite directions, and smaller for AT4, TA6 and AT7. As the chemical shift is dominated by ring current effects of neighbouring

T A B L E VI P R O P E R T I E S O F T H E I M I N O P R O T O N S IN T H E trp H O L O R E P R E S S O R - O P E R A T O R C O M P L E X k~Sis the rate constant for exchange at 3 5 ° C in the absence of the repressor, and k 35 is the rate constant in the presence of the repressor. Values in parentheses were calculated from the measured activation parameters, n.d., not determined. Base

CG1 GC2 TA3 AT4 TA6 AT7 GC5/8 TA9 TA10

8(free)

8(bound)

A8

-- 10 3 dS/dt ( p p m / o C)

k~5

k 3n5

(ppm)

(ppm)

(ppm)

free

bound

(s - 1)

(s ~ 1)

-- 13 12.75 13.42 13.37 13.52 13.49 12.6 13.72 13.44

13.08 12.72 13.44 13.44 13.52/13.63 13.63/13.52 12.58 13.98 13.20

n.d. - 0.03 + 0.02 + 0.073 0 / + 0.11 0.16/0.03 - 0.02 + 0.26 - 0.24

n.d. 1.25 2.8 3.0 2.9 2.8 1.22 3.3 2.8

5.3 4.6 8.8 8.8 8.8 7.8 1.0 10.6 n.d.

n.d. (1052) (578) (123) 35 35 7 16 21

16 (43) 142 142 36 48 24 43 22

69 base-pairs, these differential, and in opposite sense, changes in chemical shift surely reflect local changes in structure, presumably combinations of changes in local pitch and helical twist angles. These data suggest that significant alterations of the structure of the operator occur in the region TA9 to AT12, which is where we have previously observed a higher degree of conformational flexibility [8]. Conclusions The tight binding of the trp repressor to the operator is apparently not predominantly driven by cation release, because the dependence of the dissociation constant on the concentration of salt is weak, and there are no large effects on the N M R parameters of the lysine and arginine residues. However, there are presumably multiple close contacts between the repressor and the major groove of the operator because of the increased relaxation of the H8's of the bases. However, because these resonances are so broad in the complex, it was not possible to identify the protein groups responsible using nuclear Overhauser effects.

Changes in the structure of the repressor The structure of the repressor is apparently not greatly perturbed in the complex, as most of the N M R spectrum can be accounted for by a 50% increase in correlation time, with the exception of the surface tyrosine residue, whose internal motion may be somewhat quenched. There are only two tyrosine residues in the trp repressor, one of which is buried, and the other is freely accessible to solvent [14,16]. According to Schevitz et al. [9], the first seven amino acids give weak electron density in the X-ray diffraction map, suggesting that the N-terminus has a flexible arm. Tyrosine-7 may therefore be the surface-exposed residue. It has also been suggested that the N-terminal arm wraps around the D N A double helix to make contact in the minor groove [9,31], analogous to the C1 repressor [32]. The observation of decreased internal motion of the tyrosine residue is consistent with this notion. The present resolution of the spectra is not sufficient, however, to give a more detailed description of the effects of the

interaction on the trp repressor. We note, though, that the repressor is a very stable protein, having a melting temperature above 70 ° C [14], whose subunits require 6 M urea to denature them [33], and a stable hydrogen bonded core [9,15]. Internal motions seem to be of significant amplitude only for residues on the surface of the molecule [15]. Further, the crystal structure of the holorepressor shows extensive intertwining of helices across the intersubunit binding surface [9]. It is possible therefore, that major changes in the structure of the repressor do not occur, but rather only readjustments of surface regions in the DNA binding site.

Operator structure The large anisotropic forces generated by the interaction are expected to have some structural effects, as the macromolecules are not rigid bodies. Indeed, the trp operator shows evidence of conformational flexibility in its pseudorotation in the deoxyriboses, and the temperature-dependent conformational changes in the Pribnow box [6,8]. Further, in the interaction of the trp repressor with dA20dT20, we were able to detect small but significant changes in the structure of the D N A moiety, involving the propeller twisting and the sugar puckers [18]. The nuclear Overhauser effects and relaxation data on the H I ' presented here (Table II) suggest that there are significant effects on the amplitudes of the pseudorotation, and possibly sugar puckers. The clearest effects, however, are seen in the chemical shifts and exchange-rate constants of the imino protons. Ignoring the effects on the three terminal base-pairs, which can be attributed to stabilisation against fraying, there are large effects on the chemical shifts and temperature-dependence of the chemical shifts in the central region TA9 to AT12 (Pribnow box), and smaller effects in the regions TA3 to G C 8 and CG13 to AT18. Such effects can be attributed to a change in the conformation of the DNA. We note that the conformation of the TTAA sequence is sensitive to the temperature, with changes in helical twist and sugar puckers occurring as the temperature is raised. Conformational changes in other parts of the D N A were not observed [8]. It seems reasonable, then, that this region of intrinsically higher conformational flexibility is the one most

70

likely to yield to anisotropic forces. The repressor is a symmetric dimer, and binds to a symmetric operator. A possible conformational change is a kink in the centre of the D N A , i.e., at TA10, A T l l . Schevitz et al. [9] have modelled the interaction of;the trp repressor with the operator assuming rigid bodies and standard B geometry for the D N A . In their model, the bihelix motif makes intimate contact with the major grooves from TA3 to GC10 and CG13 to AT18, with the subunit interface opposite the minor groove consisting of TA9 to AT12. This arrangement would be expected to exert a bending force on the D N A , producing a kink at the centre of the D N A . Such a kink may require changes in local pitch, twist angle and tilt of the bases in the Pribnow box, and therefore considerably affect the chemical shifts of the imino protons. Accommodating changes in the backbone would affect the crankshaft motions believed to drive the local fluctuations responsible for exchange of the imino protons [30]. We note also that similar observations have been observed for the interaction of the cro repressor with OR3 [33,35]. The N M R and CD data were interpreted as a possible localised conformational change in the D N A , with only minor changes in the structure of the repressor. A more detailed description must await further model-building studies taking into account the sequence dependence of the structure of the operator, its conformational flexibility, and the results presented in this paper. Such studies are in progress. Acknowledgements This work was supported by N I G M S Grant No. GM 33385. J.F.L. gratefully acknowledges N A T O and the Phillipe Foundation for grants. We thank Drs. R.W. King and A. Aulabaugh for reading the manuscript and their comments. References 10hlendorf, D.H. and Matthews, B.W. (1983) Annu. Rev. Biophys. Bioeng. 12, 259-284 2 Carruthers, M.H. (1980) Aces. Chem. Res. 13, 155-160 3 Calladine, C.R. (1982) J. Mol. Biol. 161, 343-352 4 Dickerson, R.E. (1982) J. Mol. Biol. 166, 419-441 5 Patel, D.J., Kozlowski, S.A. and Bhatt, R. (1983) Prec. Natl. Acad. Sci. USA 80, 3908-3912 6 Jardetzky, O., Lane, A.N., Lef~vre, J-F., Lichtarge, O., Hayes-Roth, B. and Buchanan, B. (1986) in 'NMR in the Life Sciences,' pp. 49-72, Plenum, New York

7 Nussinov, R. (1985) J. Theor. Biol. 115, 179-189 8 Lefrvre, J-F., Lane, A.N. and Jardetzky, O. (1985) FEBS Lett. 190, 37-40 9 Schevitz, R.W., Otwinowski, Z., Joachimiak, A., Lawson, C.L. and Sigler, P.B. (1985) Nature 317, 782-786 10 Matthew, J.B. and Ohlendorf, D.H. (1984) J. Biol. Chem. 260, 5860-5862 11 Weber, I. and Steitz, T.A. (1984) Prec. Natl. Acad. Sci. USA 81, 3973-3977 12 Frederick, C.A., Grable, J., Melia, M., Samudzi, C., JenJacobson, L., Wang, B.-C., Greene, P., Boyer, H.W. and Rosenberg, J.M. (1984) Nature 309, 327-330 13 McClarin, J.A., Frederick, C.A., Wang, B-C., Greene, P., Boyer, H.W., Grable, J. and Rosenberg, J.M. (1987) Science 234, 1526-1541 14 Lane, A.N. and Jardetzky, O. (1985) Eur. J. Biochem. 152, 395-404 15 Lane, A.N. and Jardetzky, O. (1985) Eur. J. Biechem. 152, 405-410 16 Lane, A.N. and Jardetzky, O. (1985) Eur. J. Biechem. 152, 411-418 17 Lane, A.N. (1986) Eur. J. Biochem. 157, 405-413 18 Lefrvre, J-F., Lane, A.N. and Jardetzky, O. (1985) J. Mol. Biol. 185, 689-699 19 Lane, A.N., Lefrvre, J-F. and Jardetzky, O. (1986) Biochim. Biophys. Acta 867, 45-56 20 Kirpichnikov, M.P., Yartzev, A.P., Minchenkova, L.E., Chernov, B.K. and lvanov, V.I. (1985) J. Biomolec. Str. Dyn. 3, 529-536 21 Bax, A. (1982) Two Dimensional NMR in Liquids, Reidel, Amsterdam 22 Wagner, G. and Wi~thrich, K. (1979) J. Mag. Res. 33, 675-679 23 Platt, T. (1980) 'The Operon' (Miller, J.H. and Reznikoff, W.S., eds.), pp. 263-302, Cold Spring Harbor Laboratory, NY 24 Joachimiak, A., Kelley, R.L., Gunsalus, R.P., Yanofsky, C. and Sigler, P.B. (1983) Prec. Natl. Acad. Sci. USA 80, 68-672. 25 Barkley, M.D. and Bourgeois, S. (1980) 'The Operon' (Miller, J.H. and Reznikoff, W.S., eds.), pp. 177-220, Cold Spring Harbor Laboratory, NY 26 Lane, A.N., Lefrvre, J-F. and Jardetzky, O. (1986) J. Mag. Res. 66, 201-218 27 Kirpichnikov, M.P., Kurechkin, A.V., Chernov, B. and Skryabin, K.G. (1984) FEBS Lett. 175, 317-320 28 Kearns, D.R. (1984) CRC Crit. Rev. Biechem. 15, 237-286 29 Chou, S.-H., Wemmer, D.E., Hare, D.R. and Reid, B.R. (1984) Biochemistry 23, 2257-2262 30 Englander, S.W. and Kallenbach, N. (1984) Q. Rev. Biophys. 16, 521-555 31 Oppenheim, D.S., Bennett, G.N. and Yanofsky, C. (1980) J. Mol. Biol. 144, 133-142 32 Pabo, C.O., Kroratin, W., Jeffery, W. and Sauer, R.T. (1982) Prec. Natl. Acad. Sci. USA 80, 2184-2188 33 Pabo, C.O. and Lewis, M. (1982) Nature 198, 443-447 34 Lane, A.N. and Jardetzky, O. (1987) Eur. J. Biechem., in the press 35 Kirpichnikov, M.P., Hahn, K.D., Buck, F., Riiterjans, H., Chernov, B.K., Kurechkin, A.V., Skryabin, K.G. and Bayev, A.A. (1984) Nucleic Acids. Res. 12, 3551-3561